<p>Accurate positioning ensures the successful completion of tasks for high-maneuverability aircraft. For radar systems, the point-by-point positioning method fails to fully integrate measurement data with the kinematic characteristics of the aircraft, making it impossible to accurately obtain both the aircraft’s trajectory and systematic errors in the measurement data. Due to the flexible maneuverability and significant trajectory variations of aircraft, this work utilizes a non-uniform B-spline to constrain the trajectory and construct a fusion positioning model based on radar measurement data. Then, the position and velocity Cramér-Rao lower bound (CRLB) of the fusion positioning model is theoretically derived, and the relationship between spline fitting accuracy and aircraft positioning accuracy is deduced, thereby highlighting the importance of spline knot optimization. Therefore, a nonlinear positioning module residual neural network (NPM-RNN) is constructed to optimize the spline knot based on range and range-rate measurement data. Redundant knots are deleted based on sparse optimization and model selection to provide initial knots for the NPM-RNN. Furthermore, the NPM-RNN transforms spline knot position optimization into neural network parameter optimization, establishing an implicit mapping relationship from initial knots to optimal knots by using radar measurement data. The NPM-RNN can simultaneously optimize the spline knot positions in the <i>x</i>, <i>y</i>, and <i>z</i> directions, ultimately obtaining the trajectory parameters of the aircraft and the systematic errors in the measurement data. The simulation results show that compared with the other 10 mainstream optimization methods, the NPM-RNN proposed in this paper can effectively optimize spline knots and obtain highly accurate trajectory parameters and systematic errors. Therefore, the proposed method can provide theoretical references for airplanes, satellite positioning, and other applications.</p>